ABSTRACT
The reliability of traditional asset pricing tests depends on: (1) the correlations between asset returns and factors; (2) the timeâseries sample size T compared to the number of assets N. For macroârisk factors, like consumption growth, (1)â(2) are often such that traditional tests cannot be trusted. We extend the GibbonsâRossâShanken statistic to test identification of risk premia and construct their 95% confidence sets. These sets are wide or unbounded when T and N are close, but show that average returns are not fully spanned by betas when T exceeds N considerably. Our findings indicate when meaningful empirical inference is feasible.
This article is protected by copyright. All rights reserved